Apply for an Embedding Key
Currently, WiseMind AI requires users to configure their own Embedding Keys. The following models are supported:
All costs generated by using the key are charged entirely by the respective model provider and have no relation to WiseMind AI.
| Model Provider | Application URL | Model List | Default Value |
|---|---|---|---|
| Tencent Hunyuan | Get API Key | Get Model List | hunyuan.tencentcloudapi.com |
| MiniMax AI | Get API Key | Get Model List | embo-01 |
| Zhipu AI | Get API Key | Get Model List | embedding-2 |
| Baidu Qianfan | Get API Key | Get Model List | embedding-v1 |
| Alibaba Tongyi | Get API Key | Get Model List | text-embedding-v1 |
| OpenAI | Get API Key | Official Website | text-embedding-3-small |
| OpenRouter | Get API Key | Official Website | openai/text-embedding-3-small |
| Gemini | Get API Key | Official Website | gemini-embedding-001 |
| Cloudflare | Get API Token | Model List | @cf/baai/bge-m3 |
| Voyage AI | Get API Key | Official Website | voyage-3.5 |
| Cohere | Get API Key | Official Website | embed-v4.0 |
| Jina AI | Get API Key | Model List | jina-embeddings-v4 |
Configure and test
Open Settings → Model Settings → Embedding Models in WiseMindAI. Select a provider, then enter the API key, model name, and endpoint. Cloudflare also requires an Account ID.
After saving, select Test on the current model card. A successful test displays the vector dimensions, confirming that the model can convert text into vectors. The test does not change your active model or create a production document index.
Cost notice
Both online tests and document analysis call the provider's service and may incur a small fee. Pricing and quotas are determined by the provider.
Recommendations
- For primarily Chinese content, choose an Embedding model that explicitly supports Chinese or multiple languages.
- If you change the Embedding model for an existing knowledge base, reprocess its documents to avoid mixing vectors from different models.
- If the test succeeds but document chat still cannot find content, make sure the document analysis has completed.
- Models differ in dimensions, price, and maximum input length. Compare a small set of documents before choosing a model for a long-term knowledge base.